Cyclin-dependent kinase 1 (CDK1) and CDK2 have opposing roles in regulating interactions of splicing factor 3B1 with chromatin

Splicing factor 3B1 (SF3B1) is a core splicing protein that stabilizes the interaction between the U2 snRNA and the branch point in the mRNA target during splicing. SF3B1 is heavily phosphorylated at its N terminus and a substrate of cyclin-dependent kinases (CDKs). Although SF3B1 phosphorylation coincides with splicing catalysis, the functional significance of SF3B1 phosphorylation is largely undefined. Here, we show that SF3B1 phosphorylation follows a dynamic pattern during cell cycle progression that depends on CDK activity. SF3B1 is known to interact with chromatin, and we found that SF3B1 maximally interacts with nucleosomes during G1/S and that this interaction requires CDK2 activity. In contrast, SF3B1 disassociates from nucleosomes at G2/M, coinciding with a peak in CDK1-mediated SF3B1 phosphorylation. Thus, CDK1 and CDK2 appear to have opposing roles in regulating SF3B1–nucleosome interactions. Importantly, these interactions were modified by the presence and phosphorylation status of linker histone H1, particularly the H1.4 isoform. Performing genome-wide analysis of SF3B1–chromatin binding in synchronized cells, we observed that SF3B1 preferentially bound exons. Differences in SF3B1 chromatin binding to specific sites, however, did not correlate with changes in RNA splicing, suggesting that the SF3B1–nucleosome interaction does not determine cell cycle–dependent changes to mRNA splicing. Our results define a cell cycle stage–specific interaction between SF3B1 and nucleosomes that is mediated by histone H1 and depends on SF3B1 phosphorylation. Importantly, this interaction does not seem to be related to SF3B1's splicing function and, rather, points toward its potential role as a chromatin modifier.

SF3B1, also known as SAP155, is a core component of the U2 small nuclear ribonucleoprotein complex (snRNP) 3 that is essential for pre-mRNA splicing. The U2 snRNP interacts with the branch point adenosine during splicing catalysis, and SF3B1 is known to stabilize the interaction between the U2 snRNA and the branch point adenosine (1,2). Whole-genome sequencing has identified recurrent SF3B1 mutations in multiple neoplastic processes. SF3B1 HEAT repeat domain mutations are most commonly found in myelodysplastic syndrome, specifically in the refractory anemia with ring sideroblasts subtype, a clonal hematopoietic disorder characterized by anemia and characteristic morphologic atypia with immature erythroid progenitors containing perinuclear, iron-laden mitochondria (3,4). SF3B1 mutations are also found in uveal melanoma, chronic lymphocytic leukemia and breast cancer, albeit at lower frequencies (3)(4)(5)(6)(7). Whereas much attention has been focused on understanding the role of SF3B1 HEAT domain mutations in disease pathogenesis (8 -12), many questions regarding the function and regulation of SF3B1 during splicing, with particular relevance to the development of therapeutic strategies for the treatment of SF3B1-related disorders, remain unanswered.
In addition to its role during splicing catalysis, SF3B1 interacts with chromatin via chromatin-remodeling proteins (13,14) and may co-regulate certain histone modifications (15). A large body of evidence suggests that RNA splicing occurs cotranscriptionally and that components of the spliceosome machinery interact with chromatin during transcription (16,17). Kfir et al. (18) recently demonstrated that SF3B1 interacts with nucleosomes near exons in an RNA-independent manner. Using genome-wide occupancy data for SF3B1 in combination with splicing analyses and knockdown approaches, the authors contend that SF3B1 occupancy of chromatin determines splicing outcomes (18). SF3B1 has long been known to be a substrate protein for cyclin-dependent kinases (CDKs), although current understanding of the function of CDK-dependent phosphorylation is relatively sparse in light of the number of phosphosites potentially utilized by these kinases (Table 1). Most known CDK sites reside within the N terminus domain of SF3B1 outside the HEAT motif-containing region, whereas relatively few reside in the C-terminal portion. SF3B1 associates with cyclin E, and both CDK1 and CDK2 are known to phosphorylate SF3B1 (19,20). SF3B1 is also phosphorylated by DYRK1A, which contributes to the regulation of pre-mRNA splicing (21,22). Phosphorylation at specific amino acid residues within the SF3B1 N terminus are known to be important in mediating its interaction with other nuclear proteins, including NIPP1 (20), and phosphorylation of SF3B1 is temporally associated with active splicing (23,24). The presence of phosphorylated SF3B1 in spliceosomes during active splicing is particularly interesting, given evidence suggesting that splicing is coordinated with cell cycle progression and follows a cell cycle stage-specific program (25). Given that CDKs are also known to interact with chromatin and phosphorylate various chromatin-associated proteins (26,27), we hypothesized that phosphorylation of SF3B1 during cell cycle progression regulates its interaction with nucleosomes and in turn could influence cell cycle stage-specific splicing of exons and introns in proximity to SF3B1-bound nucleosomes.
In this paper, we report that SF3B1 phosphorylation is dynamic during cell cycle progression and dependent on CDK activity. The cell cycle-dependent phosphorylation of SF3B1 differentially regulates its interaction with nucleosomes during G 1 /S and G 2 /M, with CDK2 and CDK1 playing opposing roles in regulating the interaction. Data from in vitro binding studies demonstrated that the interaction between SF3B1 and mononucleosomes depends on the presence and phosphorylation status of linker histone H1. Using chromatin immunoprecipitation followed by high-throughput sequencing (ChIP-Seq) and RNA-Seq of synchronized cells, we mapped regions of the genome of cell cycle stagespecific occupancy of SF3B1 within genes and determined the extent to which these correlated with changes in splicing. Our findings in toto provide new insights into the role of phosphorylation of SF3B1 by CDKs and suggest that splicing-independent functions may be regulated by cell cycle-dependent SF3B1-chromatin binding.

SF3B1 phosphorylation is dynamic during cell cycle progression
SF3B1 contains numerous serine or threonine residues juxtaposed to a proline at the ϩ1 position, representing potential sites for phosphorylation by proline-directed kinases, including cyclin-CDKs, glycogen synthase kinase 3 (GSK3), and mitogenactivated protein kinases. Although several MS studies have demonstrated SF3B1 phosphorylation sites clustering within the N-terminal domain of the protein (Table 1), the functional significance of SF3B1 phosphorylation is largely undefined. SF3B1 phosphorylation is coupled with splicing catalysis (23); specifically, phosphorylation of threonine 313, a cyclin E-CDK2 substrate, is associated with active splicing (24). Cyclin E-CDK2-mediated phosphorylation of SF3B1 at threonines 244, 248, and 313 has been shown to mediate interaction with NIPP1, a protein phosphatase that localizes within the nucleus (20). To understand first whether SF3B1 phosphorylation changes during cell cycle progression, we used mitotic arrest to synchronize two human cell lines, HeLa and K562, and measured SF3B1 expression and serine/threonine phosphorylation (using anti-phosphoserine/phosphothreonine-proline (pSer/ Thr-Pro) and anti-phosphothreonine 313 (pSF3B1) (24) antibodies by immunoblot. We found a highly dynamic pattern of SF3B1 phosphorylation through cell cycle progression, wherein SF3B1 is phosphorylated at low levels in G 1 , declining further at G 1 /S, but is highly phosphorylated at G 2 /M (Fig. 1, A and B). Based on previous MS data that demonstrated CDK-mediated phosphorylation of SF3B1, we hypothesized that the increased phosphorylation of SF3B1 in G 2 /M depended on CDK1 activity. We further hypothesized that the subsequent decrease in SF3B1 phosphorylation following mitotic exit was mediated by phosphatases active during reentry into G 1 . To test these hypotheses, we first treated HeLa and K562 cells arrested in G 2 /M with purvalanol A, a specific CDK1 (CDC2) inhibitor (28,29). Brief pharmacologic inhibition of CDK1 (at 0.5 and 4 h) completely blocked detectable SF3B1 phosphorylation (Fig.  1, C and D). Inhibition of phosphatase activity in G 1 /S cells using okadaic acid (OA) treatment that inhibits both PP2A Table 1 Proline-directed serine/threonine CDK substrate motifs in SF3B1 Shown are the proline-directed serine/threonine phosphorylation sites in SF3B1 listed by the PhosphositePlus tool (53) and previously identified by MS. All serine and threonine residues are shown in lowercase letters. Predicted CDK phosphosites have been underlined. Six threonine residues were mutated to alanine to understand their impact on SF3B1-nucleosome interactions and are denoted by boldface, italicized text.

CDKs regulate SF3B1-chromatin interactions
and PP1 phosphatases (30) resulted in increased SF3B1 phosphorylation (Fig. 1E). Together, these results demonstrate that SF3B1 phosphorylation is dynamic during cell cycle progression, peaking at G 2 /M in a CDK-dependent manner and then decreasing due to phosphatase activity as cells progress through G 1 /S.

SF3B1-nucleosome interactions are dynamic during cell cycle progression
Work by several groups has demonstrated that SF3B1 interacts with chromatin or chromatin-associated proteins (13,14). Of note, Kfir et al. (18) showed that SF3B1 associates with mononucleosomes near exons and positively influences splicing of these occupied exons. Moreover, splicing is known to be coordinated with cell cycle progression, such that the expression and splicing of specific transcripts are regulated in a stagespecific manner (25). In light of these data and our finding of dynamic phosphorylation of SF3B1, we hypothesized that changes to phosphorylation of SF3B1 during cell cycle progression influence its association with nucleosomes. To test this, we first prepared whole-cell lysates and mononucleosome-enriched fractions from mitotically arrested and synchronized HeLa and K562 cells. Mononucleosome fractions containing nuclear proteins and chromatin were prepared by isolating nuclei and digesting them with micrococcal nuclease (MNase), such that the resulting DNA fragments were ϳ150 bp in size (length of DNA wrapped around a single nucleosome is 147 bp) ( Fig. 2A, inset). We compared SF3B1 protein abundance in nucleosome-enriched versus whole cell lysates of synchronized cells. In nucleosome-enriched lysates, SF3B1 levels are significantly reduced in G 2 /M, when SF3B1 phosphorylation peaks ( Fig. 1), and increased in G 1 and G 1 /S. However, SF3B1 protein abundance remained unchanged between G 1 , G 1 /S, and G 2 /M in whole-cell lysates (Fig. 2, A and B), supporting our hypothesis that cell cycle-dependent phosphorylation influences SF3B1 association with nucleosomes.
Next, we asked whether the change in chromatin association was specific to SF3B1 or if other spliceosome proteins also exhibit cell cycle-dependent chromatin association. We hence determined the total abundance and nucleosome association of SF3B2, a component of the U2 snRNP (similar to SF3B1), and also of U1-70K (or snRNP70), a subunit of a distinct snRNP (U1). Both SF3B2 and U1-70K followed the same cell cycledependent chromatin association as SF3B1 (Fig. 2C). Next, to confirm that this cell cycle-related change in splicing factor abundance within chromatin-enriched lysates was directly linked to a change in interaction with nucleosomes, we examined the association of SF3B1, SF3B2, and U1-70K with the core nucleosome protein, histone H3, in synchronized cells. As expected, SF3B1, SF3B2, and U1-70K did not associate with histone H3 in G 2 /M cells but showed increased histone association in G 1 /S (Fig. 2, D and E). Treatment of G 1 /S HeLa cells with OA resulted in decreased SF3B1-nucleosome interaction (Fig. 2F). Because OA treatment did not alter cell cycle distribution (not shown), this result suggests that the dynamics of SF3B1 phosphorylation and not merely the progression of cells through G 1 /S regulate SF3B1-nucleosome interactions.
Splicing is a co-transcriptional process; hence, it is possible that the observed cell cycle-dependent interaction between spliceosome proteins and chromatin are dependent on nascent, transcribed RNA or snRNA within the spliceosome. To address this, we tested the effect of RNase on the interaction of the spliceosome proteins with histone H3 and chromatin. Whereas RNase-A treatment significantly diminished the interaction between histone H3 and SF3B2 and U1-70K in G 1 /S phase, the SF3B1 interaction with nucleosomes remained intact, consistent with an RNA-independent mode of association during G 1 /S (Fig. 2G), consistent with previous reports (18).
Taken together, our data demonstrate that SF3B1 interactions with nucleosomes are dynamic during cell cycle progression. SF3B1 interacts with nucleosomes during G 1 and G 1 /S stages, where it demonstrates lower level phosphorylation, whereas its interaction with nucleosomes is greatly diminished during G 2 /M when CDK1 activity peaks, suggesting that SF3B1 phosphorylation during G 2 /M directly contributes to its dissociation from nucleosomes. Importantly, unlike other spliceosomal proteins we tested (both U2 snRNP components and non-U2 components), this cell cycle-dependent interaction of SF3B1 with histone is RNA-independent and hence probably not solely due to transcription.

SF3B1-chromatin interactions are dependent on CDK activity
We next hypothesized that dynamic SF3B1-nucleosome interactions are dependent on CDK-dependent phosphorylation. We tested this hypothesis by pharmacologic inhibition of CDK1 activity in G 2 /M and of CDK2 activity in G 1 /S. Treatment of G 2 /M cells with purvalanol A restored the interaction between SF3B1 and nucleosomes (Fig. 3A). In contrast, inhibition of CDK2 activity in G 1 /S by treatment with roscovitine (31) resulted in diminished SF3B1-nucleosome interaction (Fig. 3, B and D), suggesting that CDK1 and CDK2 play opposing roles in regulating SF3B1-nucleosome interactions during cell cycle progression. Another kinase, DYRK1A, is also active during G 1 and is known to phosphorylate SF3B1 at threonine 434 (21). We hence evaluated the role of DYRK1A in regulating SF3B1nucleosome interactions. Inhibition of DYRK1A by treatment with a selective inhibitor, harmine (32), did not affect SF3B1nucleosome interaction, whereas the inhibitor did elicit an increase in cyclin D2 protein, which is normally destabilized in response to DYRK1A-mediated phosphorylation (33) (Fig. 3, C and D). These data suggest that CDK1 and CDK2 play a selective role in regulating SF3B1-nucleosome interactions during cell cycle progression.
Previous studies have implicated specific threonine residues within the N-terminal region of SF3B1 in mediating its interaction with other proteins, including NIPP1 (20). Whereas our data from pharmacologic inhibition of kinase activity addressed . Immunoblot analysis to assay SF3B1 phosphorylation was performed using an antibody that detects phosphorylation at all serine/ threonine residues with a proline at the ϩ1 position (pSer/Thr-Pro) and a site-specific antibody that detects phosphorylation at threonine 313 of SF3B1 (Thr-313). The bottom panels in A and B show flow cytometric analysis of cells synchronized and harvested for analysis in G 2 /M, G 1 , and G 1 /S phases. Cells were labeled with propidium iodide (PI) to measure DNA content and model cell cycle. Histogram peaks representing diploid and tetraploid DNA content are labeled as 2N and 4N, respectively. G 2 /M-arrested HeLa (C) and K562 cells (D) were treated with purvalanol A (Purv-A) for the indicated times, and SF3B1 was immunoprecipitated from whole-cell lysates followed by immunoblot analysis to assay SF3B1 phosphorylation. E, G 1 /S HeLa cells were treated with 20 nM OA for 12 h, and SF3B1 was immunoprecipitated from whole-cell lysates. SF3B1 phosphorylation was determined as shown.

CDKs regulate SF3B1-chromatin interactions
the role of overall phosphorylation, the role of phosphorylation of SF3B1 at specific amino acid residues in regulating its interaction with nucleosomes is unclear. MS has identified CDK-dependent phosphorylation of SF3B1 at multiple sites in addition to those CDK sites previously described as regulating the NIPP1 interaction, including threonines 142, 211, 257, 261, 426, and 434, all of which have ϩ1 prolines (34). We hence mutated all six of these MS-identified threonine residues (Table 1) (21,34). On probing the SF3B1-nucleosome interactions in K562 cells expressing WT or the compound phosphomutant SF3B1 (6A), we found partially decreased nucleosome interaction of 6A, suggesting that aggregate phosphorylation of SF3B1's N terminus is required for regulation of SF3B1-nucleosome interactions (Fig. 3E). Consistent with this result, we found only a CDKs regulate SF3B1-chromatin interactions modest decrease in cyclin E-CDK2-mediated phosphorylation of 6A when compared with WT ( Fig. S1), probably reflective of the large number of N-terminal SF3B1 phosphosites contributing to overall phosphorylation of the protein. Indeed, single point mutations in SF3B1 phosphosites showed no alteration in nucleosome interactions, whereas a more extensive mutant containing 25 alanine substitutions within the predicted S(P/ T)P phosphosites was not expressed sufficiently in cells to study (data not shown).

SF3B1-chromatin interactions in vitro depend on linker histone H1 and phosphorylation of both SF3B1 and H1
A large number of nucleosome-binding proteins influence chromatin organization. SF3B1 is part of the SF3B complex, which in turn forms part of a large multiprotein complex, the U2 snRNP, that is also known to contain chromatin-associated proteins (35). It is thus unclear whether the interaction between SF3B1 and nucleosomes is direct or depends on other proteins known to complex with nucleosomes. To help determine which of these scenarios is more likely in vivo, we reconstituted SF3B1-nucleosome binding in vitro. Given that the majority of serine/threonine residues identified as potential substrates for CDKs in SF3B1 are located within an N-terminal domain that excludes the HEAT-containing domain (Table 1), we subcloned this fragment (aa 1-500, or SF3B1(1-500)) for use in these studies. Examining previously published MS-based SF3B1 interactome data sets, we noted that histone H1 is part of the complex that contains SF3B1 and other nucleosome proteins (18,36). H1 is involved in chromatin organization during cell cycle progression and is a known CDK substrate during cell cycle progression (37). Using co-immunoprecipitation, we first confirmed that SF3B1, histone H1, and histone H3 are in complex in vivo (Fig. 4A). The SF3B1(1-500) with N-terminal GST tag was then expressed in Escherichia coli and isolated for in vitro binding assays (Fig. 4B). We confirmed that SF3B1(1-500) could be phosphorylated by purified cyclin E-CDK2 in vitro (Fig. 4C). Using an in vitro binding assay in which SF3B1(1-500) is first phosphorylated in vitro using purified cyclin E-CDK2 and then incubated with purified HeLa mononucleosomes in the presence or absence of purified calf thymus histone H1, also phosphorylated in vitro by cyclin E-CDK2, we found that phosphorylation of both histone H1 and SF3B1(1-500) was required for robust binding of the latter to purified mononucleosomes. However, prolonged SF3B1(1-500) incubation with cyclin E-CDK2 in vitro led to reduced binding (Fig.  4D, top), suggesting that lower amounts of SF3B1 phosphory-lation are permissive to chromatin interactions, whereas hyperphosphorylated SF3B1 impedes the interaction.
Histone H1 has multiple isoforms; histones H1.1, H1.2, H1.3, H1.4, and H1.5 are ubiquitously expressed in a cell cycle-dependent manner, whereas histone H1.0 and the H1.X isoforms are expressed mainly in differentiated cells independent of cell cycle (37). Among the ubiquitous isoforms, H1.4 undergoes phosphorylation during S phase and mitosis. H1.4 phosphorylation has been detected at serine residues 172 and 187 in interphase cells, and additional phosphorylations at serine 27 and threonines 18, 146, and 154 have been detected in mitotic cells (38). Importantly, H1.4 is known to maintain S-phase progression, as selective depletion of it leads to a decrease in cell cycling and S phase (39). Also, with affinity purification and MS, H1.4 was recently identified as a specific interacting partner of SF3B1 (36). We thus hypothesized that H1.4 promotes SF3B1nucleosome interactions during G 1 /S. To test this, we performed separate in vitro binding assays of SF3B1-nucleosome interactions using either native calf thymus histone H1 isoform mix, purified human histone H1.4, or histone H1.0. As shown in Fig. 4E (top), we found that the presence of purified histone H1.4 most enhances the interaction between SF3B1(1-500) and mononucleosomes compared with the other linker histone preparations. This result suggests that the dynamics of linker histone isoform interactions with mononucleosomes during cell cycle progression contribute to the regulation of SF3B1chromatin interactions.
Moreover, we found that phosphorylation of H1.4 by CDK2 results in the most robust SF3B1-mononucleosome interaction in vitro (Fig. 4E). Notably, H1 in complex with H3 undergoes an increase in phosphorylation from G 1 /S to G 2 /M in vivo (Fig. 4F), suggesting that whereas H1.4 phosphorylation promotes SF3B1(1-500) interaction with mononucleosomes in vitro, high levels of phosphorylation on both SF3B1 and H1, evident during G 2 /M, diminish SF3B1-nucleosome interactions. In this way, phosphorylation of both SF3B1 and linker histone H1 may enable switchlike regulation of U2-chromatin interaction during cell cycle progression, by permitting histone binding when both are phosphorylated at lower levels (e.g. during G 1 /S) and disfavoring binding when they are hyperphosphorylated during progression to G 2 /M.

Integrated analysis of SF3B1 genome occupancy and cell cycle-dependent splicing
Our biochemical data demonstrate that SF3B1-nucleosome interactions are regulated by CDK-dependent phosphorylation Figure 2. SF3B1-nucleosome interactions are dynamic during cell cycle progression. SF3B1 protein abundance was assayed in nucleosome-enriched and whole-cell lysates from HeLa (A) and K562 cells (B) synchronized in G 1 , G 1 /S, and G 2 /M. Histone H3 (H3) and ␤-actin were used as nuclear and whole-cell lysate loading controls, respectively. Inset, DNA was isolated from MNase-digested nuclear lysates using phenol/chloroform extraction and run on an agarose gel to assess the size of the DNA fragments. C, total protein abundance of SF3B2 and U1-70K was assayed in nucleosome-enriched and whole-cell lysates of HeLa cells synchronized in G 1 /S and G 2 /M. U1-70K Santa Cruz Biotechnology sc-390988 and SF3B2 Abcam ab56800 antibodies were used in this experiment. H3 and ␤-actin were used as nuclear and whole-cell lysate loading controls, respectively. D, H3 was immunoprecipitated from nucleosome-enriched lysates of HeLa cells synchronized in G 1 /S and G 2 /M. SF3B1 co-immunoprecipitation was assayed by immunoblot assay. E, H3 was immunoprecipitated from nucleosomeenriched lysates from HeLa cells synchronized in G 1 /S and G 2 /M. SF3B2 and U1-70K co-immunoprecipitation was assayed by immunoblot using the U1-70K antibody from Dr. Doug Black's laboratory and SF3B2 (Novus 79848) antibody. F, G 1 /S HeLa cells were treated with 20 nM OA for 12 h, and H3 was immunoprecipitated from nucleosome-enriched lysates. SF3B1 co-IP was assayed by an immunoblot assay. SF3B1 in co-IP was quantified and normalized to immunoprecipitated histone H3. The experiment was performed in triplicate, relative SF3B1 abundance in co-IP was quantified using NIH ImageJ, and S.D. was calculated. G, H3 was immunoprecipitated from nucleosome-enriched lysates of HeLa cells synchronized in G 1 /S. SF3B1, SF3B2, and U1-70K co-immunoprecipitation was assayed after RNase A treatment of the immunoprecipitation reaction as shown. The immunoblot displayed is representative of three independent assays.

CDKs regulate SF3B1-chromatin interactions
of SF3B1 and linker histone H1. To understand the functional significance of this SF3B1-nucleosome interaction (and specifically how this interaction relates to stage-specific splicing pro-grams), we performed SF3B1 ChIP followed by sequencing (ChIP-Seq) and paired-end RNA sequencing (RNA-Seq) using G 1 /S and G 2 /M HeLa cells. Mononucleosome-enriched frac- Bottom, roscovitine-mediated inhibition of CDK2 activity was confirmed by analysis of cyclin E autophosphorylation (54) using conditions similar to those in B. C, top, H3 was immunoprecipitated from G 1 /S HeLa cells treated with DYRK1a inhibitor harmine (HRM) for 12 h. SF3B1 co-immunoprecipitation was assessed by immunoblot assay. Bottom, DYRK1a inhibition was confirmed by analysis of cyclin D2 protein levels in whole-cell lysates of HRM-treated G 1 /S HeLa cells by immunoblot assay. D, SF3B1 abundance in H3 co-IPs from drug-treated relative to vehicle-treated cells was quantified from 3-4 experiments and is displayed with means (thick bars) and S.D. values (thin bars). E, FLAG-tagged WT and compound phosphosite mutant SF3B1 (6A) were overexpressed in K562 cells by retroviral transduction. Cells were synchronized in G 1 /S, and H3 was immunoprecipitated from nucleosome-enriched lysates. SF3B1 co-immunoprecipitation was assayed by immunoblot analysis using an antibody against the FLAG tag in triplicate experiments.

CDKs regulate SF3B1-chromatin interactions
tions from synchronized HeLa cells (G 1 /S and G 2 /M, in biological duplicates) were immunoprecipitated with antibodies against SF3B1. Illumina compatible libraries were prepared from DNA fragments that immunoprecipitated with SF3B1, sequenced, and analyzed per informatics pipeline detailed under "Experimental procedures." Input libraries were pre-

CDKs regulate SF3B1-chromatin interactions
pared before immunoprecipitation (IP). Control libraries (nonspecific antibody control and input) were also prepared. Resulting reads were preprocessed and mapped to the hg19 genome assembly using STAR Aligner. Correlation coefficients of biological replicate samples were confirmed before further downstream analysis (Fig. S2). Peak calls and comparative analysis were performed using HOMER (40). G 1 /S cells had higher specific SF3B1-binding peaks relative to G 2 /M (7386 versus 4096) (Files S1 and S2) compared with input controls. We also found that in G 1 /S cells, SF3B1 peaks were overrepresented within exons, compared with G 2 /M cells, in which SF3B1 was found to occupy primarily intergenic sites (Fig. 5A). Similar to a previous observation by Kfir et al. (18), the density of peaks in exons when normalized to proportional length in the human genome was significantly higher in exons than those in introns or intergenic areas in G 1 /S, but not G 2 /M cells. (Fig. 5B). A modest enrichment of exonic reads was noted relative to surrounding intronic regions in both G 1 /S and G 2 /M (Fig. S3). G 1 /S and G 2 /M peaks were largely distinct, with only 167 (1.45%) peaks overlapping between the two data sets (File S3). G 1 /S-specific peaks also showed a distribution skewed toward intragenic regions, within both exons and introns (Fig.  5C), consistent with our observations for total peaks in G 1 /S. When analyzed for DNA motifs, both G 1 /S and G 2 /M data sets showed highly significant enrichment of nonoverlapping DNA motifs (Fig. S4). In summary, our ChIP-Seq results confirm our biochemical data showing increased affinity of SF3B1 for chromatin during G 1 /S compared with G 2 /M. Importantly, the increased binding in G 1 /S is primarily in transcribed intragenic areas (including transcription start and end sites).
Next, paired-end RNA-Seq was performed using matched samples from synchronized HeLa cells that were used for the ChIP-Seq analysis. cDNA libraries were prepared from poly(A)-enriched total RNA, and paired-end sequencing was performed (75 bp, average depth of ϳ122 million). Reads were aligned to hg19 using STAR aligner and first analyzed for changes in gene expression by Cufflinks (41). In line with previous reports, significant differences were found in gene expression profiles of G 1 /S and G 2 /M (1248 differentially expressed genes enriched in cell cycle and cell death/apoptosis pathways; Files 4 and 5). We then used the rMATS (42) algorithm to iden-tify alternative splicing events in G 1 /S and G 2 /M samples. A total of 48,105 alternative splicing events (distributed across five different types: alternative 3Ј splice site (A3SS), alternative 5Ј splice site (A5SS), retained intron (RI), mutually exclusive exon (MXE), and skipped exon (SE)) were detected between G 1 /S and G 2 /M, of which 2340 met the statistical cut-off (File S6 and Fig. 5D, FDR Ͻ 0.05 and ⌬PSI Ͼ5%). Our results are in agreement with previous reports that suggested specific changes to RNA transcriptome linked to change in cell cycle (25).
To test our hypothesis that SF3B1-chromatin occupancy affects cell cycle stage-specific transcriptomes by influencing splicing and/or gene expression in proximity to the site of gene occupancy, we determined overlap between genes enriched in three different data sets: G 1 /S-specific chromatin binding (ChIP-Seq), gene expression (Cufflinks), and alternative splicing (rMATS analysis). To determine whether SF3B1 occupancy of chromatin positively influences transcription of genes in those regions or changes alternative splicing, we determined overlap of genes co-localized with G 1 /S-specific peaks with transcripts or alternative splicing events increased in G 1 /S compared with G 2 /M. In contrast to our expectations, as shown in Fig. 5E, there appeared to be little overlap among these gene sets, not reaching statistical significance by Fisher's exact test.
To examine the possibility that SF3B1 binding of nucleosomes decreased transcription or alternative splicing, a similar analysis was performed for transcripts and splicing events that decreased in G 1 /S compared with G 2 /M (Fig. S5). Little overlap was evident in this analysis as well. Importantly, no G 2 /M peaks were found to be specific when compared with G 1 /S when analyzed with similar HOMER parameters. Taken together, our genome-wide integrative analyses suggest that change in SF3B1-chromatin interactions during cell cycle progression do not reflect accompanying changes to transcription or splicing of genes at the same loci. (19,20), and SF3B1 phosphorylation is known to be both associated with active spliceosomes and important for mediating protein-protein interactions (23,24). Our data now provide a new function for SF3B1  1-500)) that is encoded by an inducible expression construct used in B-F is shown in relationship to the HEAT motif-containing region spanning most of the protein (amino acids 463-1304) (55). Bottom, N-terminal GST-tagged SF3B1(1-500) expression in E. coli after IPTG induction was assessed by immunoblot analysis using an antibody that recognizes an N-terminal region of SF3B1 (Abcam, ab172634). C, SF3B1(1-500) expressed in E. coli was isolated from whole-cell lysates after IPTG induction and phosphorylated in vitro using purified cyclin E-CDK2 for the indicated incubation times. SF3B1 phosphorylation was assessed by immunoblot analysis. D, top, N-terminal GST-SF3B1 expressed in E. coli was phosphorylated in vitro with purified cyclin E-CDK2 for the indicated times and incubated in vitro with purified mononucleosomes in the presence or absence of phosphorylated native calf thymus histone H1. ϫ, slower migrating species of SF3B1(1-500) with 30-min cyclin E-CDK2 incubation, compared with faster migrating species E without phosphorylation in vitro. N-terminal GST-SF3B1-mononucleosome interactions were determined by immunoblot analysis. Bottom, equal levels of total H3 and H1 and phosphorylated histone H1 added to the in vitro binding reaction were confirmed by immunoblot analysis. Histone H1 (Active Motif, 39707) antibody was used, and H1 phosphorylation was assessed using the phosphoserine/phosphothreonine-proline (Abcam, ab9344) antibody. A representative result of three independent replicates is shown. E, top, N-terminal GST-SF3B1 expressed in E. coli was phosphorylated in vitro with purified cyclin E-CDK2 for 15 min and incubated in vitro with purified mononucleosomes in the presence of -phosphatase-treated or cyclin E-CDK2-phosphorylated native calf thymus H1 containing a mix of isoforms (H1m), purified human histone H1.0, or purified human histone H1.4. Bottom, immunoblot showing levels of H3, total H1, and phosphorylated H1. Mononucleosomes and linker histone H1 (H1m, H1.0, or H1.4) were added to the binding assays, as shown in equal quantities by mass. Data displayed are representative of two independent experiments. F, H3 was immunoprecipitated from nucleosome-enriched lysates of G 1 /S-and G 2 /M-synchronized HeLa cells. Abundance and phosphorylation of co-immunoprecipitated histone H1.4 was assessed by immunoblot analysis. Antibodies recognizing H1.4 specifically (histone H1.4 (Abcam ab105522)) and phosphorylated H1 (phosphohistone H1 (Abcam, ab4270)) were used. A representative result of three independent replicates is shown.

CDKs regulate SF3B1-chromatin interactions
phosphorylation by CDKs in regulating SF3B1-nucleosome interactions. SF3B1 phosphorylation at proline-directed serine/threonine sites peaks at G 2 /M, which causes it to dissociate from nucleosomes. Consistent with the varied functions of cyclin B/CDK1 in mitosis initiation, we speculate that CDK1dependent SF3B1 disassociation from chromatin may be nec- Lengths of peaks determined to be in exon, intron, or intergenic regions were normalized to the total length of these regions in the hg19 genome build. Values for G 1 /S and G 2 /M ae plotted side by side. C, differential enrichment of peaks (G 1 /S versus G 2 /M). A total of 456 peaks were found to be differentially enriched in G 1 /S compared with G 2 /M peaks using the HOMER algorithm. Their relative distribution across genomic regions is plotted. D, analysis of altered splicing in G 1 /S versus G 2 /M. Using the rMATS algorithm, aligned RNA-Seq files were analyzed for altered splicing. Distribution of 2340 events that met statistical cut-off (FDR Ͻ 0.05 and ⌬PSI Ͼ 5%) across five different types of events (alternative 3Ј splice site (A3SS), alternative 5Ј splice site (A5SS), retained intron (RI), mutually exclusive exon (MXE), and skipped exon (SE)) are shown on the left. Distribution of these events (overrepresented in G 1 /S or G 2 /M) is shown on the right. E, overlap of genes between rMATS, Cuffdiff, and HOMER analysis of G 1 /S and G 2 /M data sets. Genes with significant changes in splicing (FDR Ͻ 0.05) in any of the five splice event types (alternative 3Ј splice site, alternative 5Ј splice site, skipped exon, mutually exclusive exon, or retained intron) with a higher isoform ratio in G 1 /S compared with G 2 /M were included in the rMATS set. The Cuffdiff set includes genes significantly overexpressed in G 1 /S compared with G 2 /M (p Ͻ 0.05). The HOMER set contains genes co-localized with intragenic peaks (290 of the total 456) enriched in G 1 /S compared with G 2 /M. Overlaps were not found to be significant by Fisher's exact test. Error bars, S.D.

CDKs regulate SF3B1-chromatin interactions
essary to ensure efficient chromatin condensation, given the size of the spliceosome, which could serve as an impediment to this process if remaining associated.
We have also learned that SF3B1-nucleosome interactions in vitro are dependent on the presence and phosphorylation of linker histone H1, especially the H1.4 isoform. Our in vivo data and previously reported MS-based studies (36) demonstrate that H1.4, which plays an important role in the regulation of chromatin organization and transcription, interacts with SF3B1. H1.4 methylation at lysine 26 promotes its binding to HP1 and L3MBTL1, leading to the recruitment of these factors, with roles in heterochromatin formation and transcriptional repression, to chromatin (43,44). H1.4 acetylation at lysine 34, on the other hand, is known to co-localize on active promoters with a transcriptional activator, TAF1, and have a positive effect on transcription (45). Similar to SF3B1, H1.4 also undergoes phosphorylation during cell cycle and contains a number of CDK substrate motifs. Phosphorylation at threonine 187, a CDK substrate motif, is associated with RNA polymerase-mediated transcription (38), and phosphorylation of H1.4 at serine 27 inhibits HP1 binding and potentially heterochromatin formation (43). Given that SF3B1 and histone H1 exist in a complex together and are both phosphorylated during cell cycle progression by CDKs and the recognized role of H1 in regulating transcription and chromatin organization, the potential for SF3B1 and histone H1 acting collaboratively on chromatin in regulating chromatin structure and transcription should be further explored. Additionally, phosphorylated H1.4 has been found in the nucleolus, suggesting a sequestration function to regulate the activity of binding partners (46). Studies utilizing high-resolution imaging to examine whether H1 isoforms interact with SF3B1 differentially during cell cycle progression and alter SF3B1 subcellular localization may provide additional mechanistic detail on the regulation of SF3B1chromatin interactions.
To understand the functional outcomes of cell cycle-dependent SF3B1-nucleosome interactions, we performed ChIP-Seq and RNA-Seq in synchronized cells. ChIP-Seq revealed a cell cycle stage-specific pattern of SF3B1 chromatin occupancy with increased binding within transcribed regions (exons, introns, and transcription start and termination sites) in G 1 /S compared G 2 /M with genome-wide analysis (Fig. 5A). Whereas RNA-Seq of matched samples revealed cell cycle stage-specific splicing and expression changes, in agreement with a previous report (25), we found no evidence that SF3B1 occupancy on chromatin within these intragenic regions influences their expression level or alternative splicing in corresponding transcripts in a cell cycle stage-specific manner. Our findings were surprising because they differ from a previous report by Kfir et al. (18) that suggested a direct, positive effect on the splicing of exons located near SF3B1-bound regions of nucleosomes. It is important to note some differences between the design of these two studies; the conclusions of Kfir et al. (18) were not based on correlation of genome occupancy and changes in splicing, but rather on a comparison of exon utilization after SF3B1 knockdown or trichostatin A treatment. Our results are based on direct correlation of SF3B1 occupancy to changes to the transcriptome in two cell cycle states with the maximal identified difference in SF3B1-nucleosome interaction. We did not observe significant alterations in cell cycle kinetics or apoptosis with overexpressing the compound phosphosite mutant (6A) SF3B1 versus WT (Fig. S6), although a relatively limited impact of these amino acid substitutions on overall phosphorylation (Fig. S1) and histone binding (Fig. 3E) may account for the absence of an obvious phenotype with overexpression of this mutant protein. Further investigations into the functional importance of SF3B1-nucleosome interactions and their possible role in regulation of transcription (independent of splicing) and downstream biology are warranted.
SF3B1 associates with a number of proteins involved in control of chromatin modifications and transcription. For example, SF3B1 interacts with NIPP1 in a phosphorylation-dependent manner, and NIPP1 is implicated in the regulation of EZH2 occupancy at promoter regions (47,48). SF3B1 and PHF5a are found together in the SF3B complex, and PHF5a has been shown to regulate RNA polymerase-dependent transcription of pluripotency genes (35). Furthermore, it has been shown that splicing promotes the recruitment of methyltransferase HYPB/Setd2 to chromatin, resulting in histone H3 lysine 36 methylation at actively transcribed intron-containing genes (15). A potential nonsplicing-related role for cell cycle-dependent SF3B1-nucleosome interactions might be to enable key chromatin modifications by facilitating the recruitment of chromatin modifying proteins and complexes to transcriptionally active chromatin. Thus, our data may point to a previously unappreciated role for CDKs: coordinators of chromatin modifications via the modulation of SF3B1-nucleosome interactions during cell cycle progression.

Cell culture
HeLa cells were cultured in Dulbecco's modified Eagle's medium (Gibco) supplemented with 10% fetal bovine serum (Gemini Bioproducts) and 1ϫ penicillin/streptomycin (Gibco). K562 cells were cultured in RPMI 1640 (Gibco) supplemented with 10% fetal bovine serum and 1ϫ penicillin/streptomycin. HeLa and K562 cells were synchronized in G 2 /M by culturing in their respective growth media with 40 ng/ml nocodazole (Sigma-Aldrich, M1404) for 20 h. For G 1 and G 1 /S synchronizations, HeLa cells were cultured in growth medium with 40 ng/ml nocodazole for 16 h. For all synchronization experiments, cycle status was verified using flow cytometry as described below. Cells were then washed with growth medium once and cultured in growth medium for 20 h (G 1 ) or 24 h (G 1 /S). K562 cells were cultured in growth medium with 40 ng/ml nocodazole for 20 h, washed once with growth medium, and cultured in growth medium for 4 h (G 1 ) or 7.5 h (G 1 /S). Pharmacologic inhibitors were added to G 1 /S-synchronized cells for the indicated times following release from nocodazole treatment at the following concentrations: roscovitine

Flow cytometry
For cell cycle analysis, cells were fixed with 75% ethanol and incubated at 4°C overnight. Cells were then washed once with 1ϫ PBS (Gibco) and stained with 10 g of propidium iodide (Sigma-Aldrich, 81845), 0.1% BSA (Sigma-Aldrich, A2153), and 10 g of RNase A (Sigma-Aldrich, R6513) in 1ϫ PBS. Flow cytometry was performed on an LSRII flow cytometer (BD Biosciences), and analysis used FlowJo software. Apoptosis detection was performed using the Annexin V apoptosis detection kit I (BD Biosciences).

Plasmids
Codon-optimized FLAG-tagged SF3B1 (11) was cloned into BamHI and EcoRI sites of pBABE-Puro and pUC57 plasmids. Threonine-to-alanine point mutations in SF3B1 were introduced by site-directed mutagenesis using the following prim-

IP and immunoblotting
Whole-cell lysates were prepared by lysing cells in radioimmune precipitation assay buffer (150 mM NaCl, 1% Nonidet P-40, 0.1% SDS, 0.5% SDS, 50 mM Tris, pH 7.5, and 2 mM EDTA) supplemented with protease and phosphatase inhibitors (10 mg/ml each of aprotinin, leupeptin, and pepstatin, 50 mM sodium fluoride, and 1 mM sodium orthovanadate). Immunoprecipitations were performed by rotating lysates at 4°C overnight with the indicated antibodies (2 g/sample) conjugated to Protein A-Sepharose (Life Technologies) or Protein G-Sepharose (GE Healthcare) beads for antibodies produced in rabbit or mouse, respectively. Beads were washed three times with radioimmune precipitation assay buffer, and equal amounts of samples were electrophoresed and transferred to nitrocellulose membranes. Immunoblots were performed using the indicated antibodies. Nucleosome-enriched lysates were prepared as described (18) with some modifications. Cells were lysed using nucleosome lysis buffer (60 mM KCl, 15 mM NaCl, 5 mM MgCl 2 , 0.1 mM EGTA, 15 mM Tris-HCl, pH 7.5, 0.2% Nonidet P-40) supplemented with 0.1 mM DTT and protease and phosphatase inhibitors. Lysates were then passed through a sucrose cushion (50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 25 mM KCl, 1.2 M sucrose) by centrifugation at 3.5 ϫ g for 15 min at 4°C. The pellet (containing the nuclei) was resuspended in the digestion buffer (50 mM Tris-HCl, pH 7.5, 4 mM MgCl 2 , 1 mM CaCl 2 , 0.32 M sucrose) with 200 units of MNase (Worthington) and incubated for 10 min at 37°C. Digestion was stopped by the addition of 1 mM each of EDTA and EGTA.
Nuclei were pelleted by centrifugation at 3.5 ϫ g for 10 min at 4°C. Pellet was resuspended in 500 l of solubilization buffer (50 mM HEPES, pH 7.6, 500 mM LiCl, 1 mM EDTA, 0.7% sodium deoxycholate, 0.1% SDS, and 1% Nonidet P-40) and rotated at 4°C for 1 h, followed by centrifugation at 12,000 ϫ g for 10 min at 4°C. The pellet was discarded, and the supernatant, the nucleosome-enriched fraction, was used for immunoprecipitations and immunoblot assays. DNA was extracted from a small aliquot of the nucleosome-enriched fraction using phenol/chloroform/isoamyl alcohol (Thermo Fisher Scientific, 15593031) and run on a 2% agarose gel to confirm the expected size of the digested DNA (147 bp and multiples thereof). For immunoprecipitations, nucleosome-enriched fractions were diluted to 1 ml in dilution buffer (0.005% SDS, 0.1% Triton X-100, 1.2 mM EDTA, 1.67 mM Tris-HCl, pH 8.0, 167 mM NaCl) and incubated with antibodies (2 g/sample) conjugated to Protein A-or G-Sepharose beads with rotation at 4°C overnight. Beads were washed with dilution buffer three times, and immunoblots were performed as described above. RNase A treatment was performed as described (18).

In vitro binding assay
N-terminal, GST-tagged SF3B1 (SF3B1(1-500)) was expressed in bacteria and bound to GSH beads as described above. Beads were then washed three times with -phosphatase wash buffer (25 mM Tris-HCl, pH 7.5, 50 mM NaCl, 2.5 mM MnCl 2 , 2 mM DTT) and incubated with -phosphatase (New England Biolabs, P0753) for 20 min according to the manufacturer's protocol. Phosphorylation by cyclin E-CDK2 in vitro was performed CDKs regulate SF3B1-chromatin interactions as described below. Next, beads were washed once with binding buffer (0.1% Triton X-100, 1.2 mM EDTA, 1.67 mM Tris-HCl, pH 8.0, 100 mM NaCl) and resuspended in 1 ml of binding buffer, followed by incubation with 1 g of purified histone H1 (Sigma-Aldrich, calf thymus), recombinant human histone H1.0 (New England Biolabs, M2501), or recombinant human histone H1.4 (MyBioSource, MBS963223) and 1 g of purified HeLa mononucleosomes (EpiCypher, 16-0002) at 4°C with rotation for 12 h. Histone H1 was treated with -phosphatase for 20 min according to the manufacturer's protocol or phosphorylated for 30 min in kinase assay buffer as described below. Beads were washed three times with binding buffer, followed by immunoblot analysis as described before.

RNA-Seq
Synchronized HeLa cells were resuspended in TRIzol (Thermo Fisher Scientific, 15596026). Total RNA was extracted from TRIzol using the manufacturer's protocol and quantified using a Qubit fluorometer (Thermo Fisher Scientific). Poly(A) mRNA was enriched using the NEBNext poly(A) mRNA isolation kit (New England Biolabs, E7490). Libraries for paired-end sequencing were prepared using the NEBNext Ultra RNA library preparation kit (New England Biolabs, E7530) and sequenced on Illumina NextSeq 500. Approximately 511 million reads passing filter were obtained with an average of ϳ122 million reads/sample. Reads were aligned to human genome hg19.

ChIP sequencing
Nucleosome-enriched lysates were prepared from synchronized HeLa cells as described above. Lysates were diluted to 1 ml with dilution buffer and incubated with 7.5 g/sample SF3B1 antibody (MBL D221-3) at 4°C with rotation overnight. Sheep anti-mouse M-280 Dynabeads (Life Technologies, 11201D) were washed and precleared by incubation with 0.5% BSA in PBS at 4°C with rotation overnight. Beads were then added to the immunoprecipitated lysates and incubated at 4°C with rotation for 4 h. Beads were washed four times with dilution buffer, magnetically isolated, and resuspended in 150 l of SDS elution buffer (1% SDS, 10 mM EDTA, 50 mM Tris, pH 8.0) and then eluted by incubation at 65°C overnight. Samples were treated with RNase A and proteinase K followed by DNA isolation using phenol/chloroform/isoamyl alcohol. Libraries for sequencing were prepared using the NEBNext Ultra DNA library preparation kit (New England Biolabs, E7370) and sequenced on an Illumina NextSeq 500 system. Approximately 393 million total reads passing filter were generated with an average of 75 million reads passing filter per sample. Reads were aligned to human genome hg19.

Bioinformatic analysis
All Illumina reads were preprocessed with Trimmomatic (49) and aligned with Bowtie2 (50) to the hg19 genome (ChIP-Seq) or STAR (51) to the hg19 transcriptome (RNA-Seq). Peak calling for individual ChIP-Seq files was performed with HOMER (40) using the respective input files as controls. For differential peak calls and overlap of peaks, biological replicates were merged and analyzed using HOMER (getDifferential-Peaks or mergePeaks functions of HOMER). Differential enrichment of reads in the exonic region and correlation between ChIP-Seq files were performed using deepTools2 (52). Differential gene expression was determined from aligned RNA-Seq files using Cufflinks (41). Analysis of differential splicing was performed using rMATS (42). Custom perl scripts were used to integrate results of HOMER, Cufflinks, and rMATS analyses. Detailed parameters for analysis and scripts will be made available on request.